Blind direction-of-arrival (DOA) estimation algorithms based on direction-vector phase shifts in a uniform linear array are investigated in this paper. Instead of using a pilot signal for training, the output of a blind beamformer is used to compute the phase difference relative to that of a second parallel beamformer. This is possible because the two sets of beamformer weights are related to each other in a specific known manner: (i) they are identical for a subarray approach and (ii) conjugated for a reversed beamformer. Using a least-squares model, these techniques are compared with another blind DOA estimator based on the constant modulus array. Computer simulations are provided to evaluate the performance of the approaches for various cochannel interference and noise conditions.
I. INTRODUCTIONRecent direction-of-arrival (DOA) estimation techniques for a uniform linear array (ULA) exploit the phase difference between the outputs of two parallel beamformers. The beamformers can be implemented using overlapping subarrays as described in [1], or they can be a conventional beamformer and the reversed array in [2] where the antenna signals of the second beamformer are in reverse position. These lowcomplexity techniques can have a high resolution because the DOA estimate is computed at the output of the beamformers where the cochannel interference has been suppressed and the noise power has been reduced. With the aid of a pilot signal, it is shown in [1] that the subarray improves the performance of conventional DOA estimators, such as the multiple signal classification (MUSIC) [3] algorithm and estimation of signal parameters via rotational invariance techniques (ESPRIT) [4]. It is shown in [2] that the reversed beamformer has better performance than the overlapping subarrays because of the following properties: (i) it uses the full antenna array and thus can suppress more cochannel interference, and (ii) the phase difference in the reversed array is an integer multiple of the unknown phase of the signal of interest. In this paper, we investigate the performance of a reversed beamformer for blind applications where a training signal is not available for use in phase-shift-based DOA estimation. Its performance is compared with a blind version of the subarray beamformer as well as the constant modulus (CM) array [5].